publishing.co.uk · Live data
The AI Book Discoverability Index
The real domains AI engines cite when readers ask for book recommendations — drawn from 0 book audits across 16 genres and 4 engines (ChatGPT, Claude, Gemini, Perplexity), then segmented so authors know where to get listed for value. Numbers trace to live citations, not model memory.
Updated continuously as new audits run — last updated 1 Jan 1970. The sources AI leans on shift month to month, so these rankings move — worth checking back.
Top of the AI bookshelf
The domains cited most often across every genre and engine — the shelf AI reaches for first.
How much influence does Reddit really have?
Reddit is the web's single most-cited domain overall — but for book recommendations specifically, AI reaches for it far less than its reputation suggests.
Read this before pitching: across 0 book audits Reddit is cited just 0 times — rank #0, under 1% of all citations. For book discovery, AI leans far harder on reference sites (Wikipedia, Goodreads), creator video (YouTube) and specialist best-of lists than on Reddit. Its web-wide “dominance” — ~40% of citations in the largest AI-citation indices — does not carry over to books.
Retailer reality-check — how big do the shops really do?
How often AI cites each bookshop. The number in brackets is the domain's rank across all 0 sources.
The big retailers cluster together — but every bookshop trails the reference, creator and best-of sources above. AI cites where it learns about books far more than where readers buy them.
The best "best-of" sites
The curation and listicle venues AI trusts most — the highest-value places for an author to be featured.
Publishers AI cites directly
Publisher domains carry real weight — their 'best of' pages and catalogue listings get pulled into recommendations.
Where to get listed, by genre
Every genre's own most-cited sources — the author's submission map. Open a genre for the full top-10 + its niche kingmakers.
Why this matters now
AI already shapes what readers find — and it’s starting to buy, not just recommend
of US Google searches now show an AI Overview, reaching 2bn+ people a month — counted as “Google”, not “AI”.
of AI-Overview searches end with no click — the recommendation lands without ever showing as “AI traffic”.
more likely to complete a purchase when shoppers use Amazon’s AI assistant — agents are moving from recommending to buying.
As AI assistants assemble a shortlist of three-to-five titles and increasingly complete the purchase, that shortlist becomes the only set a reader sees. If your book isn’t among the sources the AI draws on, it isn’t in the shortlist — and that’s exactly what this index maps, genre by genre.
Sources: Google AI Overviews reach (Advanced Web Ranking / Xponent21, 2026) · zero-click rate on AI-Overview searches (Bain & Co; Search Engine Land, 2026) · Amazon Rufus purchase-completion uplift (Amazon Q3 2025 earnings) · agentic-checkout launches (OpenAI, Perplexity, Google, Amazon, 2025-26).
Authors: free check → £29.99 report → £99 Get Listed Playbook · Publishers: catalogue-wide audit — which of your titles are invisible to AI, imprint by imprint.
Share these findings
Free to republish with attribution — copy the embed snippet and the live card updates as the index grows.
The Amazon reality-check — the headline
Every genre's kingmaker
Books invisible to AI
Reliably recommended
The Reddit reality-check
Every genre has its own live card too — you’ll find it on that genre’s page (swap the slug, e.g. embed/genre-romance).
Embed snippet (includes the required attribution link):
<a href="https://visibility.publishing.co.uk/book-index">
<img src="https://visibility.publishing.co.uk/book-index/embed/invisible"
alt="publishing.co.uk AI Book Discoverability Index" width="600" />
</a>
<p>Source: <a href="https://visibility.publishing.co.uk/book-index">publishing.co.uk
AI Book Discoverability Index</a></p>Methodology: real cited domains aggregated live from 0 audited books across 16 genres, mined from the four citation-transparent engines — ChatGPT, Claude, Gemini and Perplexity. Runs probing the simulated Amazon assistant are excluded from this index (the paid audit also probes that surface; it never feeds this dataset). A domain must be cited for at least two different audited books to rank, and any domain whose genre citations come ≥70% from a single title is excluded from rankings. Updated continuously as new audits run. Full detail: methodology.